Anthropic, the artificial intelligence company known for its Claude chatbot, has made a significant move into climate action by joining Frontier, the carbon removal collective. The startup contributed to a new $915 million tranche of funding, bringing Frontier's total pledged capital to $1.8 billion. This makes Anthropic the first pure AI startup to join the coalition, a milestone that highlights the growing intersection between the energy-hungry AI sector and the nascent carbon removal industry.
Frontier was founded in 2022 by Stripe, Google, Shopify, and other tech companies to accelerate the development of carbon removal technologies. The coalition vets carbon removal startups and signs advance purchase agreements, providing them with critical early revenue. So far, Frontier has contracted nearly $700 million across more than 50 projects to remove 1.8 million metric tons of carbon dioxide. These credits allow member companies to offset emissions they cannot eliminate, such as those from air travel or supply chains.
Anthropic's Climate Pivot
Anthropic's membership is notable because it comes at a time when AI companies have been on an energy buying spree. The computational demands of training and deploying large language models require vast amounts of electricity, and not all of that power comes from clean sources. Anthropic, like many of its peers, has yet to publish a sustainability report. The company has previously stated it favors an "all of the above" energy strategy, which in practice can include significant purchases from polluting sources. By joining Frontier, Anthropic is signaling a willingness to address its carbon footprint, even if only marginally for now.
The move also mark's Anthropic's first climate-related deal. While Google, a founding member of Frontier, has long invested in carbon removal, Anthropic represents a new category of members: pure-play AI startups. This could encourage other AI companies, such as OpenAI or Cohere, to follow suit. The AI industry is under growing scrutiny from regulators and environmental groups for its rising energy consumption. According to a 2024 report by the International Energy Agency, data centers, which include AI training facilities, already account for about 1% of global electricity demand, a figure that could double by 2026.
A Deeper Look at Carbon Removal
Carbon removal technologies are considered essential by the UN Intergovernmental Panel on Climate Change (IPCC) for achieving net-zero emissions. They encompass a variety of methods, including direct air capture (DAC), enhanced rock weathering, bio-oil injection, ocean alkalinity enhancement, and bioenergy with carbon capture and storage (BECCS). Frontier has supported all of these approaches since its launch, spreading its bets across the portfolio in hopes that one or more will scale to deliver gigaton-level removals annually.
However, the carbon removal market remains small and expensive. Current costs for direct air capture, for example, range from $600 to over $1,000 per ton of CO2 removed, far above the price of a typical carbon offset from nature-based solutions like tree planting. Frontier's funding model aims to drive down costs by providing a guaranteed buyer. The new $915 million tranche, however, comes with increased scrutiny. Frontier announced it will fund fewer projects, focusing on those with the best chance of removing a gigaton of CO2 per year or more. New contracts will run eight to ten years, a longer timeline than previous deals, which were often shorter-term bets on early-stage startups.
This shift mirrors what is happening at Microsoft, the world's largest corporate buyer of carbon removal credits. Microsoft has signed deals with companies like Climeworks and Heirloom Carbon, but it has also made clear that it expects the industry to eventually attract government support. Frontier echoed this sentiment, stating that any new contract must show a path to government subsidy. The implication is that corporate dollars alone cannot sustain the market indefinitely, and that policy intervention—such as a carbon price or direct government procurement—will be necessary.
AI's Energy Dilemma
The timing of Anthropic's membership is critical. AI companies are racing to build ever-larger models, each generation requiring more computing power. For example, training GPT-4 is estimated to have consumed around 50 gigawatt-hours of electricity—enough to power 4,600 average U.S. homes for a year. Inference, or the process of using a model to generate responses, adds even more demand. As AI becomes embedded in search engines, coding assistants, and creative tools, its energy footprint will grow substantially.
Some AI companies have turned to renewable energy certificates (RECs) or power purchase agreements (PPAs) to offset their electricity use. But critics argue that these instruments often fail to add new clean capacity to the grid. Carbon removal credits, by contrast, directly fund the extraction of CO2 from the atmosphere, though they are still criticized as a form of "license to pollute" if used without aggressive emission reduction efforts. Anthropic's choice to join Frontier suggests it is aware of these criticisms and is taking a concrete step, albeit one that covers only a small portion of its likely emissions.
The Frontier coalition itself acknowledges the limitations. Since its launch, it has contracted only 1.8 million tons of removal, while corporate pledges total $1.8 billion. That gap between contracts and pledges indicates that many members are waiting for cheaper, more proven technologies before writing large checks. The new $915 million tranche is designed to bridge that gap, but whether it will be enough to reach gigaton scale remains uncertain.
Notably, Frontier's approach has evolved. Initially, it made many small bets to support a diversity of technologies and researchers. Now, it is concentrating its money on a few promising projects. This mirrors the broader venture capital trend in climate tech: after a boom of early-stage investments, investors are now looking for scale and commercial viability. Startups that cannot demonstrate a clear path to government subsidy or a drastic cost reduction may not receive further funding.
As the climate continues to warm, pressure on both governments and corporations intensifies. The IPCC has warned that without widespread deployment of carbon removal, even the most aggressive emission cuts will be insufficient to meet the Paris Agreement targets. Yet the public remains largely unaware of these technologies, and political appetite for subsidizing them is low. Frontier's hope is that by funding high-potential projects now, it can demonstrate commercial viability and spur government action by 2040 or sooner.
Anthropic's entry into Frontier is a small but symbolic step. It signals that even the most cutting-edge tech companies—those whose core business depends on massive energy consumption—are beginning to account for their environmental impact. Whether this translates into meaningful emission reductions across the AI industry will depend on whether other startups follow Anthropic's lead, and whether governments step up to regulate and support the carbon removal market before it's too late.
In the announcement of the new funding, Frontier emphasized that its focus will be on quality over quantity. It will fund fewer projects but with larger amounts of money and longer contract durations. The organization has also stated that it will contract as far out as 2040, implying a long-term commitment. Exactly what happens after 2040 remains unspecified, but the hope is that by then, carbon removal will have become a standard part of the global economy, supported by stable government policies and mature markets.
Source:TechCrunch News